HK1107451B - Voltage source converter - Google Patents

Description

Voltage source converter

Technical Field

The present invention relates to a Voltage Source Converter (VSC) comprising a plurality of self-commutating semiconducting elements. More precisely, the invention relates to a method and an apparatus for controlling a VSC by means of a modulation signal, such as a Pulse Width Modulation (PWM) signal. In particular, the invention relates to a VSC comprising a plurality of series connected semiconductor elements, and in particular to a converter station of a High Voltage Direct Current (HVDC) transmission line comprising such a VSC.

The PWM controlled converter described herein includes an inverter and a rectifier. Such converters may be used in low power applications, such as motor drive systems, and in high power applications, such as High Voltage Direct Current (HVDC) transmission systems and static var compensator (STATCOM) systems.

Background

A converter, more precisely a voltage source converter, provides an electrical coupling between a DC voltage system and an AC voltage system comprising one or more phases. Depending on the power direction, the converter has the function of a rectifier, which delivers electric power from the AC system to the DC system, or the function of an inverter, which delivers electric power from the DC system to the AC system. For example, the converter may be used for variable speed control of synchronous or asynchronous rotating electrical machines and High Voltage Direct Current (HVDC) long distance transmission.

The simplest converter comprises a two-stage bridge consisting of two valves. Each valve includes a single or multiple switches. Thus, the three-phase converter comprises a bridge with six valves, each of which comprises at least one switch. The switch comprises a turn-off device and a diode connected in anti-parallel therewith. By this means, the current is controllably interrupted in one direction and freely passed in the opposite direction. For high voltage applications, each valve comprises a plurality of switches in series with such shut-off devices and anti-parallel diodes.

Since the load is of the inductive type, it is necessary to arrange a diode, called "free-wheeling diode", in parallel with the switch in order to allow the load current to flow when the corresponding switch is open. The two-stage converter is further developed into a three-stage converter requiring six additional diodes. The converter is also referred to as a Neutral Point Clamped (NPC) converter bridge.

Using one bridge of a two-stage converter as an example of the AC output voltage of the converter, the amplitude, phase angle, frequency and harmonic distortion of the fundamental frequency is controlled by alternately switching on or off two valves connected to the bridge of the same phase. Thus, the AC current is controlled as desired. The pulse signal for controlling the switches is generated according to a selected Pulse Width Modulation (PWM) method.

There are many PWM methods. The most common methods are carrier based PWM, such as sinusoidal pulse width modulation, SPWM, and carrierless PWM, such as optimum pulse width modulation, OPWM. The prior art modulation technique is based on the assumption that the switching elements of the converter operate in an ideal manner, i.e. they are just turning on and off at the instant of the control instruction. In the following, this is considered to be an ideal switching instant. In practice, however, the converter output voltage waveform deviates from the waveform controlling the time of the initial instruction.

The first reason is that the switching device is not ideal. The switching device has a delayed reaction to its control signal when switching on and off, respectively. The reaction of this delay depends on the type of semiconductor, its rated current and rated voltage, the control waveform on the gate electrode, the device temperature, and in particular the actual current to be switched.

The second reason is the blanking time, or "dead time", which must be inserted between the opening (off) command of the first valve and the closing (on) command of the second valve on the same bridge. The existence of the blanking time is such that the two valves of the converter bridge are never closed at the same time in order to prevent a short circuit.

A third cause of distortion of the output voltage is the difference in the rise and fall rates dv/dt of the voltage across the switching device during off and on. This is due to the presence of parasitic capacitances in the snubber circuit or the diode. This distortion is noticeable, especially when the switching current is low.

For the reasons mentioned there will be a delay between the switching command and the actual switching event. In order to achieve an actual switching event corresponding to an ideal switching instant, the switching command must be issued in advance. Therefore, the valve operating time must be taken into account for each opening and closing. In the following, the actuation time of a valve is defined as the time difference between the actual switching command and its actual switching event. Thus, the action time includes a delayed reaction of the switching device, a blanking time, and a variation due to a low voltage rising and falling rate (dv/dt). The consequence of these parameter variations is a non-linear error between the required voltage and the true converter output voltage. This not only leads to additional low harmonics, e.g. 5 th and 7 th harmonics, but sometimes also to instability problems of the control system. Therefore, various efforts have been made to correct or compensate for these errors.

A method for processing PWM waves and devices is therefore previously known from US 5,991,176. The purpose of this method is to reduce or eliminate the effect of blanking time (also referred to as dead time) in the inverter or controlled rectifier. The known inverter is controlled by a modulator and a discriminator. The modulator functions to generate a fixed wave and the discriminator can split the wave into a plurality of waves for individual control of the individual switches. The purpose of the discriminator is to introduce a delay in closing the corresponding switch so that when a command is given to close one switch, it is always certain that the opposite switch has been opened.

The known method suggests using two modified control fixed signals, one for the case where the current is the output current and the other for the case where the current is the input current. The direction of current flow in the load will determine whether one or the other of the two modified fixed signals is to be used. The switching commands are therefore compensated for the blanking time.

A method for adaptively compensating the dead time of inverters and converters is previously known from US6,535,402. The aim of the method is to compensate for the effect of dead time to avoid current distortion and torque ripple in motors driven by such inverters. This document recognizes the difficulty of measuring the zero crossing of the current and therefore suggests applying a bias current to the distorted current. Then it is determined when the current passes through the bias level of the current. A second dead time compensation is obtained from the current passing through the bias level and added to the first dead time compensation of the PWM signal.

The known prior art methods for correcting the error between the required voltage and the real output voltage are based on current measurements. The known method is therefore based on the measurement of current switches. A feed forward type of compensation is provided that only corrects for average voltage errors due to blanking time or low dv/dt. Errors due to delayed reaction of the switching devices are not taken into account. There is no feedback control or acknowledgement to determine whether the switching device is on or off at the very moment of the control command. In addition, the method described in US6,535,402 requires additional hardware components, which can be very expensive in high power applications.

The methods known in the prior art work well under certain conditions. They may not work properly under other conditions. One condition is that it results in high current ripple when the switching frequency is low and the inductance is also low. Typically, in high power applications of STATCOM and HVDC, the converter is directly connected to the grid. In this case, they will have a high switching current ripple. It is clear that in this case the direction of the current is different from one switching instant to the next.

The predicted current can be used at the next switching instant to estimate the action time of the next switch in advance. However, it is difficult to guarantee the correctness of the predicted current, since the correctness of the predicted current depends not only on the accuracy of the converter reference voltage, the measured current and the measured voltage, but also on the calculation speed of the control process.

In high power applications, such as HVDC and STATCOM, low order harmonics result in very high cost of the filtering equipment. Therefore, there is a need for a new control method that enables high precision switching, thereby eliminating the effect of the above-mentioned errors on the voltage source converter in high power applications.

Disclosure of Invention

It is a main object of the present invention to provide a method and an apparatus for controlling a voltage source converter by which the accuracy of the switching control is improved and the influence of the errors in question is minimized. It is a second object of the present invention to provide a method and apparatus that eliminates low order harmonics, such as the 5 th and 7 th harmonics, and instability problems in system control. Another object is to determine the valve actuation time with high accuracy. It is also an object to provide a method suitable for converters with high current ripple, as in power supply systems for high power applications, and converters with low current ripple, as in drive systems and other applications. It is also an object to provide a method which does not require additional hardware and does not depend on whether information from current or voltage measurements is used.

These objects are achieved by an apparatus having the features of independent claim 1, a method having the steps of independent claim 7 or a computer program having the features of independent claim 10. Preferred embodiments are described in the dependent claims.

According to the invention, the actual switching event is detected and the action time is adjusted by comparing the ideal switching instant with the detected actual switching event. The time of the ideal switching instant is subtracted from the time of the actual switching event and added to the current action time to form the adjusted action time. Thus, if the calculated difference is positive, the action time is increased, and if the difference is negative, the action time is decreased. If there is no difference between the ideal switching instant and the actual switching event, no adjustment of the actuation time is required.

The time difference between the ideal switching instant from the first pulse and the actual switching event can be used to correct the action time of the next pulse. Doing so causes two factors to be considered. First, the hardware performance for calculating the difference and the required adjustment between two adjacent pulses is huge. Again, the switching condition of the first pulse may not be the same as the second pulse. Therefore, the action time may be different, and the adjustment may be inferior to the adjustment made by calculating only the instant at which the switching instruction is transmitted.

According to the invention, the adjusted action time for a selected pulse in a first period of the fundamental frequency is used to correct the actual switching order for the same pulse in the next period of the fundamental frequency. Thus, the information obtained from the first cycle is used to determine the switching order in the next cycle. By storing the action time for a pulse in a first period of the fundamental frequency, there will be enough time to calculate the switching order adjustment for the next period of the fundamental frequency. Thus, the need for hardware performance is reduced. Since the switching conditions will be the same for corresponding pulses in adjacent periods of the fundamental frequency, variations in the delay response and operating conditions thereof involving the components are taken into account by adjusting the action time of the same pulse in adjacent periods.

According to a first aspect of the invention, these objects are achieved by a method of controlling a VSC from a PWM pulse signal comprising ideal switching instants for each switching pulse, the method comprising: detecting an actual switching event for a selected switching pulse in a first period of the fundamental frequency, adjusting an action time for the selected switching pulse by comparing an ideal switching instant with the actual switching event, correcting a switching order for a corresponding pulse in a next period of the fundamental frequency from the adjusted action time.

The operating conditions are in principle the same for each corresponding pulse in adjacent periods of the fundamental frequency. The current load is the same and the position in the cycle is the same. Thus, the reaction time for two corresponding pulses in different cycles should also be the same. By this adaptive approach, the uncertainty in determining the reaction time of the semiconductor due to current operating conditions is self-adjusted. The method can be applied to steady-state systems and also to frequency-variable systems, especially when the variation is slow.

In a preferred embodiment of the invention, for each pulse in a period, an average value of the action times is calculated from the action times of the equivalent pulses in the preceding period. The stored value is thus the average of the previous value and the new value. The calculation method is a linear average or an exponential average method.

In a further preferred embodiment of the invention, the determination of the switching event is estimated from a voltage measurement on the electrodes of the semiconductor element.

By adjusting the actual command instant for each pulse within the fundamental frequency period based on the information of the same pulse in the previous period, the voltage change over the valve will occur exactly at the moment of the control demand. The benefit of the present invention is that the low order harmonics are reduced to a minimum level. This will greatly reduce the cost of the filter. Another benefit is that control instability of the inverter with OPWM is avoided.

In a second aspect of the invention the object is achieved by a control device providing a Pulse Width Modulated (PWM) signal for controlling valves of a converter bridge. The control device comprises sensing means for detecting actual switching events of the semiconductor device, and computer means comprising memory means for calculating and storing an action time for each pulse in a period of the fundamental frequency, and actual switching instructions for correcting the corresponding pulse in the next period of the fundamental frequency. Furthermore, the device comprises signaling means for generating and transmitting information between the computer means, the detecting means and the semiconductor element in the converter. In a preferred embodiment of the invention, the PWM is carrierless PWM, e.g. optimum pulse width modulation OPWM, or carrier based PWM, e.g. sinusoidal pulse width modulation SPWM.

In a third aspect of the invention the objects are achieved by a computer program product comprising instructions for an apparatus to perform a method of correcting an actual instruction instant of a pulse in a period of a fundamental frequency using information from an equivalent pulse in a preceding period of the fundamental frequency. The computer program also calculates the action time of each switching pulse.

Drawings

Other features and advantages of the present invention will become more readily apparent to those of ordinary skill in the art from the following detailed description taken in conjunction with the accompanying drawings. Wherein:

figure 1a is a diagrammatic representation of a converter,

figure 1b is a diagrammatic representation of a bridge of a two-stage converter,

figure 2 is a graph showing an ideal pulse and the corresponding pulses applied to the upper and lower valves and the resulting voltage,

figure 3 is a schematic diagram showing different disconnect behaviors,

figure 4 is a schematic diagram showing the current ripple,

figure 5 shows the phase leg of a high voltage converter circuit,

figure 6 is a schematic diagram showing the delay of a switching event as a function of current,

figure 7 is a schematic diagram of voltage detection of a switching event,

figure 8 is a schematic diagram of delayed current detection of a switching event,

FIG. 9 is a block diagram of a first embodiment of a control method and apparatus according to the present invention, an

Fig. 10 is a block diagram of a second embodiment of a control method and apparatus according to the present invention.

Detailed Description

The bridge of the two-stage converter is shown as an example in fig. 1. Fig. 1a depicts an all three phase forced commutation bridge and fig. 1b is a phase section of the bridge. The bridge section includes a first valve V1 and a second valve V2, and has a lower DC terminal U_dnAnd an upper DC terminal U_dp. Each valve comprises at least one switching device comprising a self-commutating semiconductor element and a diode element connected in anti-parallel therewith. In the embodiment shown the self-commutating semiconductor elements comprise IGBTs. The bridge has an AC terminal U with an AC current i_ac。

When operating the converter, a blanking time, or "dead time", must be inserted between the opening (off) command of the first valve and the closing (on) command of the second valve, or vice versa. This is because the two valves of the converter bridge should never be closed at the same time, thus preventing a short circuit. The effect of the blanking time is shown in fig. 2. The first waveform 1 is an ideal switching pulse. The second waveform 2 is a command pulse applied to the first valve V1, and the third waveform 3 is a command pulse applied to the second valve V2. The fourth waveform 4 is the resulting voltage U_ac。t_bIndicating the blanking time. It is shown in fig. 2 that the phase position and the voltage time area determining the amplitude differ from the ideal pulse of the required output voltage.

As shown, a positive current value is defined as the input current. If the current is positive, the IGBT in the second valve V2 and the diode in the first valve V1 will conduct current. In such a case, the current in and the voltage across the second valve V2 will change almost immediately when its gate unit receives an open command. However, when an open command is sent to the first valve V1, the current in and voltage across the first valve V1 will not change. Only when the second valve V2 receives a close command does the current and voltage on the first valve V1 change. Therefore, the voltage at the AC terminal is different from the voltage required for control. This is shown by comparing the waveform of the ideal pulse 1 with the resulting voltage 4 of the AC terminal voltage.

If the current is negative, the diode in the second valve V2 and the IGBT in the first valve V1 will conduct current. When the open command is sent to the second valve V2, a voltage error will be generated, resulting in an AC terminal voltage as shown by the fifth waveform in fig. 2.

When the current magnitude is low, the current direction may be different from one switching event to the next. Then, during the disconnection, the diodes in both the first valve V1 and the second valve V2 may conduct current, i.e. when the first valve V1 is disconnected, the current is negative and when the second valve V2 is disconnected, the current becomes positive. In such a case, assuming that the switching device has ideal switching behavior, the AC terminal voltage will be as shown in the sixth waveform 6 in fig. 2. During the turn-off, the IGBTs in both the first valve V1 and the second valve V2 may also conduct current. In this case, assuming that the switching device has an ideal switching behavior, the AC terminal voltage will be as shown in the seventh waveform 7 in fig. 2.

It is therefore evident that the phase position and voltage time region are different from the output voltage commanded by the control when the current amplitude is large. If the current amplitude is small, the phase position may be different from the command, but the voltage time area appears to be the same as the output voltage required for control. It should be noted, however, that at low current turn-off, the voltage increases more slowly than at high current turn-off. As an example, fig. 3 shows different switching currents and their corresponding voltages on the valve during the switching off process. The voltage derivative at a switching current of 100A is significantly lower than the voltage derivative at a switching current of 2500A. Low voltage derivatives also lead to voltage errors compared to the control commanded voltage.

The switching is affected by the non-linear characteristics of the semiconductor element with respect to the switching current. These conditions are rarely the same for two adjacent pulses, especially for high power applications with low switching frequencies. Thus, the switching times of the semiconductor elements will not be the same for two adjacent pulses. The action time, which includes the blanking time in addition to the switching time, is affected accordingly. This means that adaptively calculating the action time of the next pulse from the information of the previous pulse will not help to increase the accuracy of the achieved switching event. The current 5 and the pulse signal 6 applied to the upper valve are plotted against time in fig. 4. It is apparent that the direction of current flow is different from one switching event to the next.

The phase leg of a high voltage converter circuit to which the invention is applicable is schematically shown in fig. 5. Typically, in a device (plant) connected to a three-phase alternating current network there are three phase legs, which share one DC capacitor 13. In a conventional manner, this comprises a plurality of series-connected power semiconductor devices 11, here in the form of IGBTs, and a so-called free-wheeling diode 12 in anti-parallel with each such device. In practice, the number of power semiconductor devices in series is much higher than shown in fig. 5.

The power semiconductor devices in series are connected to a DC capacitor 13, while the phase terminals 14 between the power semiconductor devices are connected via a phase reactor 15, for example a phase of an alternating voltage network. In fig. 5, the power semiconductor devices together with the diodes arranged on the phase terminal 14 form an IGBT valve, while those below constitute the other IGBT valves.

All power semiconductor devices in the IGBT valves are closed simultaneously by a signal from each schematically indicated drive unit 16, so that the power semiconductor devices in the first IGBT valve are conducting when a positive potential is desired on the phase terminal 14, and the power semiconductor devices in the second IGBT valve are conducting when a negative potential is desired on the phase terminal 14.

By controlling the power semiconductor devices according to the determined Pulse Width Modulation (PWM) pattern, the direct voltage over the DC capacitor 13 can be used to generate a voltage at the phase terminal 14, the basic component of which is an alternating voltage with the desired amplitude, frequency and phase position. This control is performed by sending control pulses from the control device 17 to the different drive units, typically via optical fibers. In fig. 5 there is a first optical fibre 9 and a second redundant optical fibre 10.

The information exchange between the control unit 17 and the drive unit 16 is a bidirectional communication via an optical fiber. Switching instructions are sent from the control unit 17 to the drive unit 16. A signal indicative of the switching event may be sent back from the drive unit 16 to the control unit 17. The control unit 17 at the low voltage potential is galvanically isolated from the drive unit 16 at the high voltage potential. An indication signal of the switching event is generated in the drive control unit.

There are a number of factors that affect the delay from the switch command to the actual switch. The switching device is not ideal and the switching characteristics are closely related to the characteristics of the gate driving unit. The switching device reacts to its control signal delay when closed and opened. The delay time depends on the semiconductor type, its current and voltage rating, the control waveform on the gate electrode, the device temperature, and in particular the actual current to be switched. The switching delay versus current is shown in fig. 6.

As shown in fig. 6, the current direction is the most important parameter. This is because different current directions will determine whether the current flows into the IGBT or into the diode at the switching instant. As already discussed before, a "dead time" or blanking time must be inserted between the closing command of the first valve and the opening command of the second valve. This blanking time dominates the current dependent switching action delay.

Due to the delayed reaction of the switching devices and the low rise and fall rates (dv/dt) of the voltage changes, the switching commands must be issued in advance to make the actual switching event occur at the ideal switching instant. However, if the actual switching event does not occur exactly at the ideal switching instant, there is a problem of inaccuracy.

A first consequence of this switching inaccuracy is that additional low harmonics, such as the 5 th and 7 th harmonics, are caused. A second consequence is that instability problems can occur in the control of the system. This is due to the non-linear error between the commanded voltage and the true converter output voltage. According to the invention, this non-linearity error is eliminated by detecting the actual switching event, estimating the time difference between the actual switching command and the actual switching event on-line, and adjusting the actual switching command of the same pulse in the next period of the fundamental frequency accordingly. This operates completely independently of the current direction and magnitude.

A first way of detecting the actual switching event is to use the measured voltage. During the disconnection, the voltage level on the electrode of one of the power semiconductor devices in the valve is measured using a voltage divider and compared with a predetermined reference value. As shown in fig. 7, the moment when the measured voltage 32 passes the reference 33 is considered as an actual switching event. At the moment of this switching event, a signal 34 is generated in the gate control unit of the semiconductor device. This signal is sent back to the valve controller to indicate the timing of the actual switching event. In case some individual semiconductor devices fail, several such signals will be sent from different semiconductor devices to their corresponding valve controllers. In the valve controller the time from sending the opening command 31 to receiving the indication of the actual switching event 34 will be stored and used for adjusting the corresponding opening command in the next cycle of the fundamental frequency.

According to a preferred embodiment, the reference voltage is approximately equal to half the steady-state voltage during the off-state.

A second way of determining the actual switching event is to use the measured current. AC current is measured and has been used for system control and protection. The measured current is sent as an input to a valve controller. For a specific type of semiconductor device with a specific gate unit and controller and a given blanking time, the relation between the switching current and the time delay from the switch-off command to the actual switch-off event can be obtained by means of a switching test. As an example, fig. 6 shows the switching current as a function of the time delay. The resulting function is established as a table or equivalent non-linear function in the valve control process. For each measured switch current, the corresponding time delay may be estimated using a table or non-linear function 41, as shown in fig. 8. The estimated time delay for each open command will be stored and will be used to adjust the corresponding open command in the next cycle of the fundamental frequency.

The general concept of a first embodiment of a control method and apparatus according to the invention is shown in fig. 9. In this embodiment, the pulse control processor PCP compensates for the delay occurring when the valve is switched by using adaptive control. Pulse t of the pulse train 19 by the drive unit comprising a valve control unit VCU_n ¹The pulse sequence is used to control the voltage source converter valve to form the fundamental frequency 18. The pulse signal 20 carrying this information is sent to a pulse control processor PCP comprised in the control device. The PCP also receives a control pulse CP representing a switching command that has been executed. The PCP calculates a pulse pair t by comparing the pulse signal 20 with the control pulse CP_n ¹I.e. how long the delay from the sent switching order to the implemented switching event is. The calculated reaction time 21 for each pulse in the fundamental frequency period is stored in a memory M.

The pwm controller, represented by the block OPWM, sends a pulse signal 22 indicative of the switching command, i.e. the desired switching command, as dictated by the system control. Representing the calculated pair pulse t_n ²Is added to the command signal 22 by adding means 24 to form a new command signal 25 in order to achieve the actual switching event at the desired instant. This new command signal 25 is sent to the control pulse generator C for implementing the switching command to the next switch.

Typically, the overall control of a converter in HVDC applications is divided into three main parts. First is the system control, which controls the active power/DC voltage and the reactive power/AC voltage and the AC current. The desired or ideal pulse is generated from the system control. Second is valve control, which corresponds to component 17 in fig. 5. Thirdly is a drive control unit, which corresponds to the component 16 in fig. 5.

The general concept of a second embodiment of the control method and apparatus according to the present invention is shown in fig. 10. In this embodiment, the pair pulse t is represented by signal 26_n ¹ToThe response time can be estimated by using the measured AC current and the function block 41 that has been described previously and shown in fig. 8. The calculated reaction time 21 for each pulse in the fundamental frequency period is stored in a memory M. Representing the calculated pair pulse t_n ²Is added to the command signal 22 by adding means 24 to form a new command signal 25, the purpose of which is to influence the actual switching event at the desired instant.

Although advantageous, the invention is not necessarily limited to the embodiments given as examples. The main idea of the invention is to use the information from one switching pulse of a first period of a harmonic period of the fundamental frequency to control the switching of an equivalent pulse in the next period. Thus, the determination of the switching event achieved may be estimated from a voltage measurement or a current measurement. Also, other modifications of the details will be apparent to those skilled in the art upon study of the teachings provided herein. Such modifications are included within the scope of the present invention.

Claims (11)

1. An arrangement for controlling a voltage source converter having a bridge (V1, V2) of at least two semiconductor self-excited commutating elements (11) connected in anti-parallel with diodes (12), the arrangement comprising means for generating a sequence of switching control pulses (19) to form a fundamental frequency (18), means for implementing a switching Command (CP) and means for detecting a switching event, characterized in that the arrangement comprises:

for calculating a reaction time (t) between a switching Command (CP) and a switching event (20) for selected pulses in a pulse sequence₁，t₂) Wherein the computer device is connected to the means for implementing a switching instruction and to the means for detecting a switching event;

and means (24) for adaptively compensating a switching command of an equivalent pulse in a next period of a harmonic period of the fundamental frequency, wherein said means for adaptively compensating a switching command is coupled to said means for generating a switching control pulse sequence and said computer means.

2. The apparatus of claim 1, wherein the means for adaptively compensating switching commands comprises a memory device for storing a calculated reaction time for each pulse in a harmonic period of the fundamental frequency.

3. The apparatus of claim 1 or 2, wherein the means for detecting a switching event comprises means for measuring a voltage on an electrode of at least one semiconductor device in a valve.

4. The apparatus of claim 1 or 2, wherein the means for generating a sequence of switching control pulses is an optimal pulse width modulator.

5. Apparatus according to claim 1 or 2, wherein said computer means for calculating the reaction time comprises means for calculating an average of the reaction time of each pulse in a harmonic period of the fundamental frequency.

6. The apparatus of claim 5, wherein the average comprises an exponential average.

7. A method for controlling a voltage source converter comprising at least two bridges (V1, V2), said bridges comprising semiconductor self-exciting commutating elements (11) each connected in anti-parallel with a diode (12), said voltage source converter further comprising a control unit (17), characterized in that,

the pulse sequence is arranged to form a base frequency,

the instant at which the switching command for the pulses of the fundamental pulse sequence is issued is defined,

the switch command is sent out, and the switch command is sent out,

the actual switching event is determined and,

comparing the actual switching event with an expected switching event, an

The instant at which the switching order of the equivalent pulse in the next harmonic period of the fundamental frequency is sent is adjusted.

8. The method of claim 7, wherein the determination of the actual switching event comprises measuring a voltage on an electrode of at least one semiconductor element.

9. The method of claim 7, wherein the adjustment of the instant at which the switching command of the equivalent pulse in the next harmonic period is sent comprises an adjustment of a blanking time.

10. A method for controlling a voltage source converter by means of a pulse width modulated pulse signal comprising information defining an ideal switching instant for each switching pulse,

determining an actual switching event for a selected switching pulse within a first period of the fundamental frequency,

adjusting the action time for the selected switching pulse by comparing the ideal switching instant with the actual switching event, an

Modifying a switching command of a corresponding pulse in a next cycle of the fundamental frequency.

11. The method of claim 10, wherein the determination of the actual switching event comprises measuring a voltage on an electrode of at least one semiconductor element.